Thanks for creating this! I see why you wouldn't want to adjudicate the question of fault but I worry that too much probability mass is on accidents like being rear-ended at a red light that there's literally nothing the car can do anything about. I think Waymo has a fair number of those. I was going to suggest a clarification that the car has to be moving, but there are plenty of ways that the car's lack of moving can be exactly what's at fault. Running markets sure is hard work!

@dreev I don't think its worth trying to assign fault when only 10-20 cars are driving. In 90 days they still shouldn't have enough miles that being the victim if someone else has driving is likely.

@WrongoPhD This strikes me as missing the factor of public sentiment. Human drivers seem willing, sometimes, to deliberately crash into entities-on-the-road they particularly dislike—see e.g. the bit on road rage here—and, while the number of Tesla robotaxis will be low, the number of human drivers in Austin who might have grudges against Musk which they could decide to take out via crashing into a Tesla robotaxi is high. (See incidents over the last few months of Tesla vandalism in the context of people being upset about DOGE; see also Austin's population generally favoring the Democratic party, voting-patterns-wise.)

None of which is to imply that it's not also possible that the Tesla robotaxis will crash at disproportionately-high rate due to their own internal issues, of course. But I don't think the external factors are negligible here.

@Tulip Oof, great point. Is it too late to add an exception for non-accidents, ie, intentional collisions?